In this work, four basins were under study: two typical shallow thermokarst lakes and two large and deep lakes located in the vicinity of Vas'kiny Dachi research station in Central Yamal (Fig. 1) within the framework of the long-term monitoring program of the Earth Cryosphere Institute (Dvornikov et al., 2016). All studied lakes are located in the area of continuous permafrost with an average ground temperature of up to -7 ∘C at the level of zero annual amplitudes and with the active layer depth varying between 0.5 and 1.3 m. The permafrost is characterized by high ice content (Dubikov, 1982). The morphometric characteristics of the studied lakes are summarized in Table 1.

The basins in the study area were mostly freshwater lakes. The predominant anion was Cl- (56.7 eq % on average among ∼ 30 lakes), with the proportion of anions in the following order: Cl- > HCO3- > SO42-. Cations were strongly dominated by Na++ K+ (58.5 eq % on average). The proportion of cations was in the following order: Na++ K+ > Mg2+ > Ca2+. The total mineralization in lakes was normally below 150 mg L-1 (Dvornikov et al., 2019). Lakes LK-002 and LK-010 are small in area, with the depth not exceeding 2.3 m (Table 1). During winter, most of the area of these lakes is frozen to the bottom. For microbial processes in these lakes, the absence of the water column stratification during the short summer season is an important factor. Thus, in deep lakes the temperatures of the surface and near-bottom layers were 15.1–15.8 and 12.0–14.2 ∘C, respectively (Table 1). Strong winds, low coasts, and the absence of the pycnocline provided for efficient mixing of the water column. In-depth analysis of the absence of stratification in the water column of Yamal lakes has been provided elsewhere (Dvornikov et al., 2019). Their bottom sediments have a high content of undecomposed organic matter. Lakes LK-003 and LK-004 are large and deep and had well-developed basins with sandy bottoms. According to their basic morphometric characteristics, we suggested lakes LK-002 and LK-010 to be the young, developing lakes of thermokarst origin, while lakes LK-003 and LK-004 were mature and of debatable origin.

Water and surface sediment samples were collected from four lakes during the period from 5 August till 9 August 2019. Water samples were collected from the entire water column at the deepest parts of the lake (Table 1). A total of 13 water samples were collected (Table 1). Water was sampled using a TD-Automatika© hydrological water sampler, dispensed into 35 mL glass vials, sealed with gas-tight rubber stoppers (avoiding gas bubbles), and covered with a perforated aluminum cap.

Bottom sediments (with a core length of up to 340 mm) were collected using a limnological stratometer with a glass tube. A total of 12 sediment samples were collected: LK-003, 0–3, 3–7, and 3–12 cm; LK-004, 0–4, 4–9, and 9–15 cm; LK-002, 0–4, 4–9, and 9–14 cm; and LK-010, 0–3, 3–6, and 6–12 cm. The thawed ground ice flowing into the lake was collected at the lower part of lake LK-004's slope (sample LK-004K). Sediment samples were then transferred into cut-off 5 mL plastic syringes, preserving the structure of the sediment core, and sealed with gas-tight rubber stoppers so that contact of the samples with air was avoided. The procedure has been described elsewhere (Pimenov and Bonch-Osmolovskaya, 2006). Samples of water and sediments were stored in a portable temperature-controlled box at +8 to +12 ∘C. Pore water from sediment samples was obtained by centrifugation at 8000 g for 10 min in a TsUM-1 centrifuge (Russia). The temperature, electrical conductivity (EC), and concentration of dissolved oxygen were measured using a WTW 3320 SET2 portable multimeter (Germany) equipped with a CellOx-325 dissolved oxygen detector and a TetraCon 325 conductometer.

Methane content in the water and sediment samples was determined using the head-space method (McAuliffe, 1971). Methane concentration was measured on a Kristall-2000M gas chromatograph (Chromatec, Russia) equipped with a flame ionization detector. The concentrations of sulfate and chloride ions were determined (after distillation and concentration) on a Staier ion chromatograph (Akvilon, Russia). Carbon content in dissolved organic matter (DOC) was determined by high-temperature incineration (non-purgeable organic carbon, NPOC) using a Shimadzu TOC-VCPH analyzer (Japan) with a measurement accuracy of ± 10 %. Average values were calculated from three samples. Statistical treatment of the results was performed using Excel 2000.

For assessment of total microbial abundance (MA) and microbial biomass, water samples in glass vials were fixed with glutaraldehyde at the final concentration of 2 %. Fixed samples (5–10 mL) were filtered through black polycarbonate 0.2 µm filters (Millipore, USA). The filters were stained with acridine orange (2 mg mL-1) (Hobbie et al., 1977) and examined under an Olympus BX 41 epifluorescence microscope equipped with an ImageScopeColor M visualization system. Cell volumes of the cocci and rods were calculated by approximating them as geometric spheres and cylinders, respectively. The cells were enumerated in 20 fields of view. Biomass was calculated using the data on the volume of microbial cells and assuming 1.0 mg mm-3 as the density of wet biomass. Specific biomass of microbial cells (B) is therefore presented in micrograms per liter (µg L-1). During microscopic examination of the stained preparations, single cells and cells associated with aggregates were enumerated separately. A group of cells with a common outline, in which the visual enumeration of individual cells was difficult or impossible, was considered an aggregate.

The rates of microbial processes – light and dark CO2 assimilation (LCA and DCA), autotrophic (hydrogenotrophic) methanogenesis (MG-h), and methane oxidation (MO) – were determined using radiotracer labeled compounds (NaH14CO3, specific activity 2.04 GBq mmol-1, Amersham, UK; 10 µCi per sample and 14CH4, specific activity 1.16 GBq mmol-1, JSC Isotope, Russia; 1 µCi per sample). To determine the LCA and DCA rates, two transparent vials and one darkened vial were used for each sampling horizon. Each transparent vial was covered with an individual sheath calibrated for the transmission of the photosynthetically active radiation corresponding to the illumination in the sampling horizon. A labeled substrate (0.2 mL as a sterile degassed water solution) was injected through the rubber stopper with a syringe. The vials were incubated for half the daylight period at in situ temperature. After incubation, they were fixed with 1 mL 0.1 N HCl and filtered through 0.2 µm nylon membranes. Photosynthetic production was calculated as the difference between the values for the dark and transparent vials. Incubation of water and sediment samples to determine the rates of other processes (MG-h and MO) was also carried out in situ. These processes (MG-h and MO) were stopped by injecting 0.5 mL of saturated KOH solution into each experimental vial (water samples) and plastic syringe (sediment samples). All experiments were performed in duplicate. After the end of the experiments, the vials were stored at 5–10 ∘C. The measurement of radioactivity of the products of microbial activity in both the experimental and control vials was performed in the laboratory as described earlier (Pimenov and Bonch-Osmolovskaya, 2006; Savvichev et al., 2018b). Radioactivity was measured on a TRI-Carb TR 2400 scintillation counter (Packard, USA). The calculation of the LCA and DCA rates was done considering the 14C-CO2 both in bacterial cells and in the extracellular dissolved organic matter. For the calculation of MO rates, 14C-CH4 conversion to CO2, biomass, and extracellular soluble OM were analyzed separately. The confidence interval for the LCA, DCA, MG-h, and MO rates varied from 10 %–40 %.

The δ13C methane value was determined on a TRACE GC (gas chromatograph; Thermo Fisher Scientific, USA) coupled with a DELTA Plus mass spectrometer (Thermo Electron Corporation, Langenselbold, Germany) using a PDB-calibrated (PDB stands for Pee Dee Belemnite) standard and calculated according to the following equation: 1δ13C=(13C/12Csample/13C/12Cstandard-1)×1000‰, where ([13C] / [12C])sample / ([13C] / [12C])standard are the ratios of occurrence of the 12C and 13C atoms in the sample and in the standard, respectively. The international PDB standard used has the isotope occurrence ratio [13C] / [12C] of 0.001172 (Craig, 1957). The error of the δ13C measurements did not exceed ± 0.1 ‰.

To collect microbial biomass, the water sample (500 mL) was passed through filters with a pore diameter of 0.22 µm. The filters were homogenized by triturating with liquid nitrogen, and the preparation of metagenomic DNA was isolated using the DNeasy PowerSoil Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The total amount of isolated DNA was about 0.5 µg per sample. The V3–V4 variable region of the prokaryotic 16S rRNA genes was obtained by polymerase chain reaction (PCR) with the primers 341F (5′-CCTAYGGGDBGCWSCAG) and 806R (5′-GGACTACNVGGGTHTCTAAT). The primers used in this work were amplified with both archaeal and bacterial sequences (Frey et al., 2016). PCR fragments were barcoded using Nextera XT Index Kit v2 (Illumina, USA). The PCR fragments were purified using Agencourt AMPure Beads (Beckman Coulter, Brea, CA, USA) and quantitated using Qubit dsDNA HS Assay Kit (Invitrogen, Carlsbad, CA, USA). Then all the amplicons were pooled together in equimolar amounts and sequenced on the Illumina MiSeq instrument (2×300 nt reads). Paired overlapping reads were merged using FLASH (Magoč and Salzberg, 2011).

The pool of 16S rRNA gene sequences was analyzed with QIIME 2 v.2019.10 (https://qiime2.org, Bolyen et al., 2019). DADA2 plugin was used for sequence quality control, denoising, and chimera filtering (Callahan et al., 2016). Operational taxonomic units (OTUs) were clustered applying the VSEARCH plugin (Rognes et al., 2016) with open-reference function using the Silva v. 132 database (Glöckner et al., 2017; Quast et al., 2013) with 97 % identity. Taxonomy assignment was performed using BLAST against Silva v. 132 database with 97 % identity.

The raw data generated from 16S rRNA gene sequencing were deposited in Sequence Read Archive (SRA) under accession numbers SRR11972844–SRP266728, available via BioProject PRJNA636944.

Phytoplankton primary production (PP) is the main process of carbon fixation in lakes. According to our data, PP in all studied lakes was relatively high (Fig. 2).

While PP values for the surface layers of lakes LK-003, LK-004, and LK-010 were similar (333–468 µg C L-1 d-1), PP in lake LK-002 was considerably higher (723 µg C L-1 d-1). In shallow lakes photosynthesis rates were high throughout the water column, including the near-bottom horizon; in deep lakes photosynthesis was not detected below 4–5 m. PP calculation for the entire water column (i.e., for the 1 m2 water column; mg C m-2 d-1) revealed integral production in deep, mature lakes to be somewhat higher than PP values for shallow lakes (Fig. 2).

DOC concentration in the water column did not vary significantly within each lake (± 0.5 mg C L-1; Fig. 3a). Large, deep lakes were characterized by slightly lower DOC (4.5–5.5 mg C L-1) than small, shallow lakes (7–8 mg C L-1). Upper horizons of lake sediments had much higher DOC (10–40 mg C L-1) and higher variability within the sediment column (Fig. 3b).

Microscopic analysis of the water column samples revealed relatively high microbial abundance (1.7–7.6 × 106 cells mL-1), typical of mesotrophic water bodies (Table 2).

Despite the strong wind conditions, which caused mixing of the water masses, MA varied slightly depending on the sampling horizon. In general, it was higher in the surface and near-bottom horizons. Both the highest MA and biomass were revealed in the near-bottom horizon of the shallow thermokarst lake LK-002, while the lowest occurred in the surface horizon of the shallow thermokarst lake LK-010. The average cell volume in the studied samples varied from 0.10–0.15 µm3, and the biomass was 220 to 930 µg L-1. The share of aggregated cells varied from 1.8 %–11 %. According to our results on total MA, all studied lakes were mesotrophic. However, this assumption is based on the data obtained during the highest seasonal warmth of the water column.

Dark CO2 assimilation (DCA) is an integral characteristic of the activity of chemoautotrophic and chemoheterotrophic microorganisms (Fig. 4).

DCA values in the surface layers of all four lakes were quite similar (14.4–25 µg C L-1 d-1), which follows from their similar hydrological and production characteristics. DCA values in the near-bottom layers varied even less (16.5–24 µg C L-1 d-1). However, the DCAint, calculated for the entire depth of the lake, shows that the values of DCAint in deep mature lakes are significantly higher than in shallow thermokarst lakes (187–217 mg C m-2 d-1 compared to 24–26 mg C m-2 d-1). High DCAint indices in deep mature lakes are determined by active heterotrophic processes that take place in all layers of the water column. It follows from the above that the depth of the lakes is an important factor determining the degree of transformation of organic matter in the particulate substance forming bottom sediments.

The concentration of dissolved methane measured in the water columns of shallow, well-mixed thermokarst lakes was relatively low (310–480 nmol CH4 L-1) with no significant differences between the surface and near-bottom horizons (Fig. 5). In deep mature lakes, methane concentrations decreased significantly (2- to 10-fold) from the near-bottom horizons (4140–975 nmol CH4 L-1) up to the surface layers (400–420 nmol CH4 L-1) (Fig. 5).

Experiments with labeled methane revealed that MO rates in the surface layer and near the bottom of the shallow, well-mixed thermokarst lakes were almost the same, although the absolute values of MO rates in the surface layers of shallow lakes LK-002 and LK-010 differed significantly (3.6 and 16.4 nmol CH4 L-1 d-1, respectively) (Fig. 6). In deeper lakes MO rates peaked in the near-bottom horizons: 14.7 and 55.3 nmol CH4 L-1 d-1 for lakes LK-004 and LK-003, respectively. MO rates decreased significantly in the upper layers of the water column correlating well with decreasing concentrations of dissolved methane (Figs. 5, 6).

Bottom sediments from the bottom surface and to the depth of 14–15 cm are described in Table 3.

The sediments of all studied lakes were similar in structure, with a loose, reddish upper horizon (to 3–4 cm depth), a denser intermediate layer of gray with some reddish interlayers, and a gray, dense lower layer. Numerous gas bubbles were visible in the sediments of lake LK-003. All samples, including the LK-004K thawed ground ice, had pelito-aleuritic texture. Methane content in the sediments varied from 10–1000 µmol CH4 dm-3 (Table 3). Rates of hydrogenotrophic (autotrophic) methanogenesis in the upper sediment layer varied within a broad range: from 3–7 to 203 nmol dm-3 d-1. The highest MG-h rate was observed in the lowermost collected sample from the deep lake LK-003, while the lowest one occurred in the lowermost horizon from the shallow thermokarst lake LK-002. The MO rate (up to 9.9 µmol dm-3 d-1) in the sediments exceeded the MG-h rates significantly. MO was more active in the dense subsurface layer than in the loose upper sediment layer. Methane carbon isotope composition (δ13C-CH4) in the sediments varied from -80.8 ‰ to -63.3 ‰. The highest content of the light carbon isotope (12C) in methane was observed in lake LK-002. In the sediments of LK-003, methane contained much more heavy carbon (13C). According to the average values of δ13C-CH4, the studied sediments form the following sequence: LK-002 – LK-004 – LK-010 – LK-003.

An analysis of the 16S rRNA gene sequences from 13 water samples and 13 sediment samples revealed the high taxonomic diversity of microbial communities in all four studied lakes (Fig. 7).

The water column microbial communities of all four lakes were rather similar. Most microorganisms belonged to the Bacteria domain (99.42 %–99.95 % of the total read number). Among them, Gammaproteobacteria (11.4 %–41.6 %), Actinobacteria (18.4 %–30.7 %), Bacteroidetes (6.2 %–24.4 %), Verrucomicrobia (5.4 %–19.1 %), Alphaproteobacteria (4.1 %–17.3 %), and Betaproteobacteria (3.2 %–16.5 %) were the most abundant groups. The share of cyanobacterial OTUs in the surface horizons was 6.7 %–12.4 %. Other bacterial taxa (phyla Acidobacteria, Chloroflexi, Firmicutes, Latescibacteria, Nitrospirae, Patescibacteria, Epsilonproteobacteria, and Deltaproteobacteria) constituted a minor part of the water column microbial communities in all four lakes, with the share of each taxon below 0.5 %.

Microbial communities of the bottom sediments differed from those of the water column. Archaea were relatively abundant in the sediments of all four lakes (20.2 %–55.6 % of the total number of reads) and were mainly represented by methanogens of the orders Methanomicrobiales, Methanosarcinales, and Methanomassiliicoccales. The upper sediment of lake LK-002 was exceptional in this respect, with the share of archaea not exceeding 3.96 % probably due to contamination of the sample with near-bottom water. This suggestion is supported by the elevated content of cyanobacteria in this sample. Bacteria of the bottom sediments of the studied lakes belonged to the phyla Chloroflexi (18.5 %–43.8 % in thermokarst lakes and 6.6 %–23.2 % in non-thermokarst ones), Bacteroidetes (4.2 %–12.6 %), Verrucomicrobia (1.8 %–9.0 %), Acidobacteria (2.3 %–6.5 %), Gammaproteobacteria (0.5 %–8.4 %), Deltaproteobacteria (2.1 %–6.1 %), Betaproteobacteria (3.2 %–7.2 % in the upper and 0.9 %–3.2 % in the lower horizons of all lakes), and Nitrospirae (1.7 %–4.8 %) in lakes LK-002 and LK-004. Other bacteria revealed in the sediments, belonging to Actinobacteria, Epsilonbacteraeota, Firmicutes, Latescibacteria, Planctomycetes, and Alphaproteobacteria, constituted less than 1.0 % of the microbial community.

The composition of the community from the thawed permafrost sample (LK-003K) differed significantly from those of the communities of both the water column and the sediments. Only in this sample were the “Ca. Woesearchaeota” OTUs (5.8 %) more numerous than Euryarchaeota OTUs (1.7 %). Members of the phylum Epsilonbacteraeota (genera Arcobacter and Sulfuricurvum) constituted 9.0 %, while their abundance in other samples did not exceed 0.65 %. The share of Deltaproteobacteria (genera Desulfuromonas and Geopsychrobacter) was 11.6 %. Desulfuromonas was not detected in the water column of the studied lakes, and its abundance in the bottom sediments did not exceed 0.81 %.

The community of the first (allochthonous) group was genetically diverse. It contained heterotrophs of several taxonomic groups which were almost absent in the samples of suspended matter and bottom sediments from all the studied lakes.

The second group (photosynthetic microorganisms) contained cyanobacteria related to the families Nostocaceae and Cyanobiaceae. In the tundra of northern Québec, Canada, phototrophic microorganisms were represented by picocyanobacterial groups Synechococcales and Nostocaceae and the nitrogen-fixing species Dolichospermum curvum (Wacklin et al., 2009), formerly known as Anabaena curva (Crevecoeur et al., 2015). In the deeper parts of Canadian lakes, the relative abundance of the OTUs of anoxygenic green bacteria of the genus Pelodictyon (Chlorobiaceae) was as high as 50 %. Our study revealed no Chlorobiaceae in either the shallow thermokarst or deep lakes.

Among heterotrophic bacteria (the third group), the dominant phyla were Actinobacteria (18.4 %–30.7 %) and Bacteroidetes (6.2 %–24.4 %), including such species as Actinobacteria “Ca. Planktophila” and “Ca. Nanopelagicus”, as well as Flavobacterium succinicans and Sediminibacterium sp. of the phylum Bacteroidetes.

Since tundra lakes are sources of atmospheric methane (Laurion et al., 2010), the composition of the water column methanotrophic community is of special interest. Because of their ability to oxidize methane, aerobic methanotrophs can significantly reduce methane emissions to the atmosphere and thus play an important role in the global methane cycle (Conrad, 1996). Aerobic methanotrophic bacteria belong to three main lineages: Gammaproteobacteria, Alphaproteobacteria, and Verrucomicrobia, with different carbon assimilation pathways (Knief, 2015). In the water column of all studied lakes, methanotrophic microorganisms were far from numerous (<0.2 %) and belonged exclusively to the order Methylococcales (class Gammaproteobacteria). Most of them belonged to the genus Methylobacter. The bacteria detected in the water samples included members of the genera Methylophilus (methanol-oxidizing bacteria), Acinetobacter (known to perform denitrification coupled with the oxidation of various organic compounds, including humic and fulvic acids) (Wen et al., 2019), and Limnohabitans, as well as “Ca. Methylopumilus turicensis”. The latter are methylotrophic planktonic psychrophilic bacteria capable of methanol and methylamine oxidation and preferring the low-temperature conditions of the hypolimnion (Salcher et al., 2015).

Gammaproteobacteria of the family Methylomonaceae were found in the sediments of all four lakes (up to 5.2 %). The sequences detected were most closely related to those of Methylobacter tundripaludum of the family Methylomonaceae (Smith and Wrighton, 2019). They were most abundant in the upper sediment layer and could account for the observed methane oxidation. Members of this family can grow under low-oxygen conditions and act as major methane consumers in stratified lakes (Oswald et al., 2017).

An analysis of the published data indicates that members of the families Methylomonaceae and Methylococcaceae predominate in many thermokarst basins (thaw ponds and lakes) in Alaska (de Jong et al., 2018; Matheus Carnevali et al., 2018) and Canada (Crevecoeur et al., 2015, 2017). Methylobacter tundripaludum is a typical component of microbial communities in Arctic bogs and swamped tundra soils (Graef et al., 2011; Liebner et al., 2009; Wartiainen et al., 2006).

A comparison of the data on MO rates and Methylobacter relative abundance indicates that these bacteria probably lay the basis for the microbial community utilizing methane in the sediments of tundra lakes. Methane oxidation can be carried out, apart from aerobic methanotrophic bacteria, by the ANME-2d group archaea assigned to “Ca. Methanoperedens”. These archaea may perform anaerobic methane oxidation coupled with the reduction of nitrate (Haroon et al., 2013) or of iron and manganese oxides (Ettwig et al., 2016; Leu et al., 2020). These organisms were detected in the sediments of all lakes except for lake LK-003, and in some cases their share was as high as 3 %. Thus, during the short warm period of the summer season, in all four tundra lakes, the microbial community of the oxidative branch of the methane cycle contained both aerobic methanotrophic and methylotrophic bacteria and ANME-2d archaea.

In the absence of sulfate ions, methanogenesis is well known to be the terminal stage of OM degradation in the sediments. Methane formation in tundra lakes (either thermokarst ones or of other types) is known to occur via either the hydrogenotrophic or the acetoclastic pathway; their combination is also possible (de Jong et al., 2018; Matheus Carnevali et al., 2015, 2018). Members of the families Methanosarcinaceae and Methanosaetaceae have been most commonly detected in archaeal methanogenic communities of the bottom sediments of thermokarst ponds. Hydrogenotrophic members of the orders Methanomicrobiales, Methanobacteriales, and Methanocellales were also detected (Matheus Carnevali et al., 2018; Negandhi et al., 2014). The predominance of members of certain taxonomic and physiological groups among the microorganisms of the methane cycle in different horizons of stratified basins has been reported for Lake Pavin, France (Biderre-Petit et al., 2019). In this meromictic lake, archaea of the genus Methanoregula predominate in the methanogenic community under anoxic conditions, while methanotrophic bacteria of the genus Methylobacter prevail both in the methanotrophic community of the oxic water column and at the oxic–anoxic interface.

The methanogenic archaeal community in the sediments of the studied Yamal lakes was represented by several OTUs of Methanoregula (order Methanomicrobiales), which constituted up to 16 % of all microorganisms. The abundance of the OTUs of Methanosaeta (order Methanosarcinales) in the sediments of LK-003 was up to 16.9 %. The abundance of the OTUs of methylotrophic Euryarchaeota of the order Methanomassiliicoccales (Dridi et al., 2012) was up to 26 % in LK-004 and up to 32 % in LK-002 sediments. Yamal sediments differed significantly in the dominance and the ratio of the major archaeal taxa from those in the sediments of the Emaiksoun thermokarst lake, Utqiagvik, Alaska, where, apart from the OTUs of Methanosaetaceae (30 %–31 %), high levels of Bathyarchaeota OTUs were revealed (17 %–24 %) (de Jong et al., 2018).

Since only the rate of hydrogenotrophic methanogenesis (MG-h) was determined in the present work, the overall MG rate is most probably an underestimate. This is indirectly indicated by our data on the high abundance of Methanosarcinales and Methanomicrobiales OTUs in the sediments of all four lakes. Members of these lineages are known to be equally capable of both acetoclastic and hydrogenotrophic methanogenesis (Crevecoeur et al., 2016). The contribution of acetoclastic methanogenesis may be rather significant. Thus, in the lake of Rotsee (Switzerland) over 90 % of methane released from the sediment into the water column was produced by acetoclastic Methanosaeta spp., with hydrogenotrophic methanogens responsible only for 7 % (Zepp Falz et al., 1999).

Thus, our calculations estimate the amount of methane produced by archaea via the hydrogenotrophic pathway in the 0–15 cm sediment layer to be in the range 0.6 to 3.6 µmol CH4 m-2 d-1 in shallow lakes and 3.5 to 14.5 µmol CH4 m-2 d-1 in deep mature lakes. The estimated amount of methane oxidized in these sediments is 310–650 µmol CH4 m-2 d-1 for shallow lakes and 300–1350 µmol CH4 m-2 d-1 for deep mature lakes. Based on the above calculations, it may be concluded that the contribution of hydrogenotrophic methanogenesis to the total methane production in the upper sediment layer does not exceed 5 %. Thus, our radioisotope studies revealed only a small fraction of methanogenesis occurring in the lakes. Methane is released from the sediments into the water column and is oxidized in the course of its diffusion from the bottom to the lake surface. The estimated amount of methane oxidized in the water column is 7.2–15.5 µmol CH4 m-2 d-1 for shallow lakes and 120–330 µmol CH4 m-2 d-1 for deep mature lakes.

The isotope composition of methane carbon (δ13C-CH4) in the bottom sediments (Table 3) correlated well with the MO rates (Fig. 8).

High MO rates were associated with elevated levels of the heavy carbon isotope (LK-003, three samples), while low MO rates were found in the sediments with a high content of the light carbon isotope (LK-002, three samples) (Fig. 8). The high content of the light carbon isotope in methane from all bottom sediments indicates its biogenic origin. Isotope analysis is, however, unable to differentiate between the modern biogenic methane and methane potentially released from thawing permafrost. MO results in the preferential consumption of the light carbon isotope, with the remaining methane enriched with the heavy isotope; this was especially evident in the case of the LK-003 sediments. At low MO rates, most of the overall methane remains unused, and the resultant methane retains high levels of light carbon (LK-002). The low MO rate in the presence of sufficient methane concentrations indicates that the process is limited by some factors other than the substrate deficiency. This is probably the result of a limited supply of oxygen or electron acceptors other than oxygen or the limited supply of biogenic elements.

Since our study was carried out in the polar summer, during the short season of the high rate of all microbial processes, comparison with the data obtained in April 2018 (Savvichev et al., 2018a) is required to understand the scale of seasonal variations. The data on the rates of microbial processes during the summer and winter seasons are presented in Table 4.

A comparison of these data shows insignificant differences between methane concentrations in the upper sediment layers of Yamal lakes in summer and winter. The differences in the rates of hydrogenotrophic methanogenesis were also insignificant. MO rates were, however, 20–50 times higher in summer. This elevated MO rate resulted in higher levels of 13C-CH4 (δ13C-CH4 to 64.9 ‰) in the residual (unconsumed) methane. The methane carbon isotope composition in winter was lighter and more homogeneous. MO rate in the water column was very low in winter, which was caused by the temperature decrease to 0–4 ∘C rather than the substrate limitation. The combination of low MO rates in winter and the formation of a thick ice cover resulted in increased methane concentrations in the water column under the ice. The highest seasonal methane concentration in the water column occurs probably immediately before the thawing of the lake ice cover. During the short summer season, phytoplankton production in the Yamal tundra lakes of various depths is relatively high (340–1200 mg C m-2 d-1). Allochthonous and autochthonous organic matter partially degrades in the water column and precipitates onto the bottom sediments where methane is the terminal product of anaerobic degradation (90–1000 µmol CH4 dm-3).

Our analysis reveals the following scheme of summer microbial processes within the water column and lake sediments of two significantly different lake types: deep mature lakes (with established stable basins) and shallow lakes (resulting from the active thermokarst on constitutional ground ice) (Fig. 9). OM decomposition occurs throughout the year. However, the MO rates in winter are low, which may result in methane accumulation within the water column. During the summer season, the major part of methane in deep mature lakes is oxidized in the water column, and low emission rates from the surface of these lakes may be expected. Most of the area of shallow lakes is frozen to the bottom during winter, and allochthonous OM remains mainly undecomposed. We therefore observed lower rates of methane production in shallow lakes during summer, but since the MO rates were also lower, we expect similar dissolved methane concentrations and methane emissions from the surface of these shallow and deep mature lakes. Given that most of the water within deep lakes remains unfrozen during winter, we might expect that while the annual methane production in these lakes is higher, their water column serves as a microbial filter for methane emission into the atmosphere.

The bottom sediments of tundra lakes are sources of methane which is of biogenic origin according to the data on its carbon isotope composition. The rates of hydrogenotrophic methanogenesis are higher in the sediments of deeper lakes than in the sediments of shallow thermokarst ones. In the sediments of both deep and shallow Yamal lakes, Methanoregula and Methanosaeta were predominant components of archaeal methanogenic communities (Fig. 9).

In the sediments of lakes LK-002 and LK-004, the percentage of methylotrophic was significant (OTUs up to 29 % and 14 %, respectively). In parallel to methanogenesis, methane oxidation occurs at rather high rates in the sediments of the studied lakes. MO is due to both ANME-2d archaea (“Ca. Methanoperedens”) and bacteria of the genus Methylobacter (OTUs up to 5.2 % of the total number of reads) (Fig. 9). Methane not consumed in the sediments is released into the water column, where it is oxidized by the community of aerobic methano- and methylotrophic bacteria, in which methylotrophic Betaproteobacteria of the family Methylophilaceae predominate. A decrease in the concentration of dissolved methane from 0.8–4.1 µmol CH4 L-1 in the near-bottom layer to 0.4 µmol CH4 L-1 at the surface is the visible geochemical result of the activity of methanotrophic bacteria in deep lakes (Fig. 9). In shallow lakes wind-mixed to the bottom, methanotrophic bacteria occur in trace amounts, their activity is low, and the geochemical methanotrophic effect is weak. Thus, the depth of a tundra lake is an important factor affecting the scale and completeness of the geochemical processes of the methane cycle caused by microbial activity. The role of the water column as a microbial filter is therefore significant.